Skip to content

Hypergraph is more powerful. Transform unstructured text into structured knowledge with LLMs. Graphs, hypergraphs, and spatio-temporal extractions — with one command.

6.6 KB
Raw Blame

General Templates

Base types and common extraction tasks.


Overview

General templates provide fundamental extraction capabilities suitable for a wide range of documents.


Base Types

base_model

Purpose: Basic structured data extraction

Output: Single structured object

Use for: Simple summaries, form-like data

=== "CLI"

```bash
he parse doc.md -t general/model -l en
```

=== "Python"

```python
ka = Template.create("general/model", "en")
result = ka.parse(doc_text)

print(result.data)
```

Output Schema:

{
    "field1": "value1",
    "field2": "value2"
}

base_list

Purpose: Ordered collection extraction

Output: List of items

Use for: Sequences, rankings

=== "CLI"

```bash
he parse doc.md -t general/list -l en
```

=== "Python"

```python
ka = Template.create("general/list", "en")
result = ka.parse(doc_text)

for item in result.data.items:
    print(f"- {item}")
```

Output Schema:

{
    "items": ["item1", "item2", "item3"]
}

base_set

Purpose: Unique collection extraction

Output: Set of unique items

Use for: Tags, categories, keywords

=== "CLI"

```bash
he parse doc.md -t general/set -l en
```

=== "Python"

```python
ka = Template.create("general/set", "en")
result = ka.parse(doc_text)

for item in result.data.items:
    print(f"- {item}")
```

Output Schema:

{
    "items": {"unique1", "unique2", "unique3"}
}

base_graph

Purpose: Basic entity-relationship extraction

Output: Graph with entities and relations

Use for: Knowledge graphs, networks

=== "CLI"

```bash
he parse doc.md -t general/graph -l en
```

=== "Python"

```python
ka = Template.create("general/graph", "en")
result = ka.parse(doc_text)

print(f"Entities: {len(result.nodes)}")
print(f"Relations: {len(result.edges)}")
```

Output Schema:

{
    "entities": [
        {"name": "Entity1", "type": "type1"},
        {"name": "Entity2", "type": "type2"}
    ],
    "relations": [
        {"source": "Entity1", "target": "Entity2", "type": "relates_to"}
    ]
}

Specialized Templates

biography_graph

Type: temporal_graph

Purpose: Extract person's life events and timeline

Best for: Biographies, profiles, memoirs

=== "CLI"

```bash
he parse tesla_bio.md -t general/biography_graph -l en
```

=== "Python"

```python
ka = Template.create("general/biography_graph", "en")
result = ka.parse(biography_text)

# Build index for interactive features
result.build_index()

# Visualize life events (with search/chat capabilities)
result.show()

# Ask questions
response = result.chat("What were the major achievements?")
print(response.content)
```

Features:

  • Extracts people, places, events
  • Captures temporal relationships
  • Timeline visualization

Example Output:

{
    "entities": [
        {"name": "Nikola Tesla", "type": "person"},
        {"name": "AC Motor", "type": "invention"}
    ],
    "relations": [
        {
            "source": "Nikola Tesla",
            "target": "AC Motor",
            "type": "invented",
            "time": "1888"
        }
    ]
}

concept_graph

Type: graph

Purpose: Extract concepts and their relationships

Best for: Research papers, technical documents, articles

=== "CLI"

```bash
he parse paper.md -t general/concept_graph -l en
```

=== "Python"

```python
ka = Template.create("general/concept_graph", "en")
result = ka.parse(paper_text)

# Get concept map
for node in result.nodes:
    print(f"Concept: {node.name}")

for edge in result.edges:
    print(f"{edge.source} → {edge.target}")
```

Features:

  • Concept identification
  • Relationship mapping
  • Hierarchical structures

doc_structure

Type: model

Purpose: Extract document structure and outline

Best for: Long documents, reports, books

=== "CLI"

```bash
he parse report.md -t general/doc_structure -l en
```

=== "Python"

```python
ka = Template.create("general/doc_structure", "en")
result = ka.parse(report_text)

print(f"Title: {result.data.title}")
print(f"Sections: {len(result.data.sections)}")

for section in result.data.sections:
    print(f"- {section.heading}")
```

Features:

  • Section identification
  • Heading hierarchy
  • Key points per section

workflow_graph

Type: graph

Purpose: Extract process workflows

Best for: Procedures, processes, workflows

=== "CLI"

```bash
he parse procedure.md -t general/workflow_graph -l en
```

=== "Python"

```python
ka = Template.create("general/workflow_graph", "en")
result = ka.parse(procedure_text)

# Build index for visualization
result.build_index()

# Visualize workflow
result.show()

# Query workflow
response = result.chat("What is the first step?")
print(response.content)
```

Features:

  • Step identification
  • Flow relationships
  • Decision points

When to Use General Templates

Use When:

  • Document doesn't fit specific domain
  • Need basic extraction
  • Exploring/document type unknown
  • Building custom workflows

Don't Use When:

  • Domain-specific template available (use that instead)
  • Need specialized fields (create custom template)

Examples

Example 1: Simple Summary

from hyperextract import Template

ka = Template.create("general/model", "en")
result = ka.parse("""
Meeting Notes:
Date: January 15, 2024
Attendees: Alice, Bob, Carol
Topic: Q1 Planning
Decision: Launch product in March
""")

print(result.data)

Example 2: Biography Analysis

ka = Template.create("general/biography_graph", "en")
result = ka.parse(biography_text)

# Build index for interactive features
result.build_index()

# Visualize life events (with search/chat capabilities)
result.show()

# Ask questions
response = result.chat("What were the major achievements?")
print(response.content)

Example 3: Research Paper

ka = Template.create("general/concept_graph", "en")
result = ka.parse(paper_text)

# Get concept map
for node in result.nodes:
    print(f"Concept: {node.name}")

for edge in result.edges:
    print(f"{edge.source} → {edge.target}")

See Also